THE JOURNAL OF CHEMICAL PHYSICS 126, 114506 (2007)

Bubble formation in lattice Boltzmann immiscible shear flow

R.S. Qin

Graduate Institute of Ferrous Technology, Pohang University of Science and Technology,

Hyojia-Dong San 31, Pohang 790-784, South Korea

Abstract - Bubble formation and the effect of shear on bubble formation in a van der

Waals fluid is investigated by means of lattice Boltzmann mesoscale simulations. In the

absence of shear, the maximum number of formed bubbles increases with undercooling

but the incubation time before bubble formation decreases dramatically. The results are in

agreement with classical phase transition theory. In shear flow, the maximum number of

bubbles is not affected by shear but the bubble growth rate is accelerated. The effect of

shear on bubble growth rate weakens at large undercoolings. The reasons are twofold. On

the one hand the highly undercooled system takes less time to complete phase transition

due to the large driving force so that there is less time to accumulate the flow effect. On

another hand the mechanism for bubble growth changes from coarsening to coalescence

at large undercoolings

I. INTRODUCTION

Bubble processes are important in beverage, food, pharmaceutical, metallurgical and

many other industries [1]. Manipulation of mesoscopic bubbles has attracted attention

because of potential applications in biological systems and electronic devices [2, 3].

However, the scientific understanding of bubble formation is in its early stages. Many
efforts have been devoted to improve the description of bubble formation within or

beyond the classical nucleation framework, for example the bubble and droplet formation

without nucleation barriers [4, 5]. Particularly in shear flow, the study of the effect of

shear on nucleation of bubbles, droplets or crystals is an open subject. The question of

whether the shear induces or suppresses nucleation is still a debating topic [6].

Experimental observations reveal versatile effects of shear on nucleation [6-8].

Observations of bubble formation are usually restricted by the resolution of analysis

equipment and the capability of sample preparation, which makes it difficult to access the

properties of tiny bubbles. Theoretical work based on kinetics has made interesting

predictions [9]. It is expected that numerical investigations can supply an independent

viewpoint and help validate theory and explain observations. Simulation such as

molecular dynamics can provide insight into the mechanism for bubble formation, but the

computational time can be prohibitive for length-scales over 1 m and time-periods in

excess of 10 ns [10]. A few coarse-graining methods have been developed for

mesoscopic simulation of phase transitions in systems where hydrodynamics are

important [11-13]. The lattice Boltzmann equation is among these mesoscopic simulation

methods [13]. The aim of the present work was to tackle bubble formation using a

mesoscopic interparticle potential developed under the lattice Boltzmann BGK

framework [14], with emphasis on the effects of undercooling and shear rate on the

maximum bubble number density, nucleation time and bubble growth rate. The system is

represented by a van der Waals equation for simplicity and yet does not lack generality

[15].
The paper is organized as follows: The theoretical approximations and related knowledge

are described in Sec. II. The simulation procedure and numerical results are discussed in

Sec. III. Finally, the work is summarized in Sec. IV.

II. THE THEORY AND APPROXIMATION

A wide class of non-ideal fluids can be approximated by a van der Waals system. The

system consists of a single phase with uniformly distributed molecules at a temperature

above the critical point Tc , but separates into two phases when the temperature falls

below Tc . When this happens, the phase with the diluted molecular density represents the

gas and that which is denser is designated the liquid [15]. For such a system, bubbles

form when the gas is the minor phase under equilibrium conditions. Otherwise, droplets

are said to form.

The mesoscopic interparticle potential in the lattice Boltzmann scheme is defined as [13,

14]

Vkl  wkl k l (1)

where k and l represent lattice positions.  is a signal function explained below. wkl is

the weight factor that depends on the relative position between position k and l. For a van

der Waals fluid  has been derived by the author according to the unification of

equations of state for both van der Waals and lattice Boltzmann fluids [14]. The van der

Waals equation of state is

 RT
p  a 2 (2)
1  b

where p, R and T are pressure, gas constant and temperature, respectively.  is the

molecular density given by   n / V where n is molar number of molecules and V is

volume of the fluid. a and b are parameters representing attraction and repulsion

between molecules. The lattice Boltzmann equation of state at the long wave limit is [16]

RT 2
p  RT   (3)
2

Identification of Eqs. (2) and (3) gives

2b 2 2a 2
 2   (4)
1   b RT

where   1 when the right hand side of Eq. (4) is negative, or   1 vice versa. The

mesoscopic interparticle potential shows strong repulsion for very large  and weak

attraction for very small . The overall shape of the interaction is similar to a Lennard-

Jones interaction, but the parameter is not distance between particles but the local

molecular density, which shows the coarse-graining effect [14].

The interaction force exerted on coarse-grained particle which is located at lattice

position k is

Fk  l k Vkl (5)

The interaction affects particle motion by changing the momentum of the particle
 
 k  vk  Fk  t (6)
where vk is the particle speed at lattice position k project to the coordinate .  vk is the

corresponding speed change.  t is the duration of a time step. In a lattice Boltzmann

scheme the coarse-grained particle is described by a set of particle distribution functions

{ f i , i=0, 1, … b}. The mass and momentum of the particle at lattice k is given by

b
k   fi (7)
i 0

 b

and  k vk   f i ci (8)
i 0


where ci is the speed vector of lattice Boltzmann model, as illustrated in figure 1.

System evolution is described by time iteration of three chronological processes named

streaming, acceleration and collision. In streaming, the particle distribution functions

propagate to their neighbors without changing their values. Particle acceleration is

computed by momentum conservation such as Eq. (6). Particle collision can be calculated

by the lattice BGK theory

fi   fi  
1

f i  f i eq  (9)

where f i and f i  are the particle distribution functions before and after collision.  is

collision time which is related to kinetic viscosity of the fluid. f i eq is the equilibrium

particle distribution function. For two-dimensional nine-neighbour regular lattice

Boltzmann model (named d2q9) as shown in figure 1, f i eq is given by [17]

     
f i eq   wi 1  3(ci  v )  4.5(ci  v ) 2  1.5 v
2
 (10)
where wi is the weight factor and takes the same value as wkl in Eq. (1). For d2q9 lattice

Boltzmann model it has w0  4 / 9 , wi  1 / 9 for i=1,2,3,4 and wi  1 / 36 for i=5,6,7,8.

One of the most advantages in using lattice Boltzmann equation over other mesoscopic

models is the easy-handling of boundary conditions. The shear boundary condition has

been developed by Zou and He [18]. For d2q9 model under boundary shear speed of

u x xˆ  u y yˆ , the unknown particle distributions on the lattice at the bottom boundary are

given by


1
 f 0  f1  f 3  2 f 4  f 7  f8  (11.1)
1 uy

2
f2  f4   uy (11.2)
3

f5  f7 
1
 f1  f 3   1  u x  1  u y (11.3)
2 2 6

f 6  f8 
1
 f1  f 3   1  u x  1  u y (11.4)
2 2 6

The unknown particle distribution on the lattice at top boundary can be obtained by

rotating Eq. (11) with 180.

III. NUMERICAL SIMULATIONS AND RESULTS

The numerical simulations in the work are performed in two dimensional nine neighbor

regular lattice Boltzmann scheme. The lattices at the top and bottom boundaries are

specified by speed boundary conditions described by Eq. (11). Periodic boundary

conditions are applied for those lattices at the left and right boundaries. Density

fluctuations are introduced through the computation of mesoscopic interactions

 k    k    (12)

where  is fluctuation intensity and  is a random number between –1 and 1. Mass

conservation is maintained throughout the simulation because no fluctuation is really

added to the molecular density. The method of introducing fluctuations has been well

used in phase-field modelling of alloy solidification [18].

To have a general concept of the length and time scales in lattice Boltzmann model, it is

supposed that the fluid is air with the kinetic viscosity of 1510-6 m2s-1 and at 100 C.

The sound speed of the lattice Boltzmann model is 327.19 m/s. The lattice distance is

158.7 nm for   1 . The time step is 2.79810-10 s. Lattices of 1024128 correspond to

dimensions of 162 m  20 m. The five-thousand time steps are equivalent to 1.4 s.

As is well-known, this is in the typical mesoscopic scale. Boundary shear speed of 0.05 at

the top and -0.05 at the bottom for fluid with thickness of 128 lattices is equivalent to a

shear rate of 2.83 106 s-1.

The van der Waals fluid considered in the present work is defined with a=1.018226 and

b=0.3. The kinetic viscosity, which is defined as the bulk viscosity divided by the fluid

density, is assumed to be constant with respect to the molecular density.   1 . The

energy unit is defined as RTc . The initial molecular distributions are uniform in all

simulations.
The number of bubbles is calculated according to following three rules. The first is that

the lattice is considered within a bubble if the molecular density on the site is less than a

critical value    c . The second is that the bubble is considered independent if it is

absolutely isolated by those nearest neighbors with    c . The third is that the

resolution for bubble is in one lattice size, which is the lowest limit in the lattice

Boltzmann scheme.

A. Effect of undercooling on bubble formation

The no shear condition is that for which the shear velocities at the top and bottom

boundary is zero. Bubble formation is investigated as a function of temperature. The

fluctuation intensity should be related to the system temperature but there is no accurate

relationship available in the lattice Boltzmann model because of the coarse-graining

procedure which smears out many of the microscopic degrees of freedom. To compensate

this, several fluctuation intensities with wide differences are considered, in the range  =

0.00001 and  = 0.005. The results for  = 0.005 are presented in Table 1.

The initial density is 0  2.0 and molecules are distributed uniformly. The state is in not

at equilibrium for the initial temperature T=0.91 ~ 0.99. The smaller initial temperature

corresponds to larger degree of undercooling. The larger undercooled system possesses

stronger thermodynamic driving forces for the formation of bubbles. The time for the

formation of the first bubble t s , designated the waiting time [10], decreases

monotonically from 440 time-steps for T=0.99 to 55 time-steps for T=0.91. In fluids that
are not at equilibrium, bubble formation and bubble coalescence coexist. Bubble

formation rate decreases with the constitutional undercooling during phase separation.

Bubble coalesce increases at larger bubble number and bubble radius. There is a time at

when the number of bubbles achieves maximum due to the competition between

formation (by nucleation) and disappearance (by coalesce) of bubbles. The time steps for

achieving the maximum number of bubbles, tmax , are smaller at larger undercoolings as

illustrated in table 1. The interval from the first formation of a bubble to the maximum

bubble stage, tmax  t s , is decreasing as the undercooling is increased. The maximum

number of bubbles increases monotonically with the undercooling, from 105 bubbles at

T=0.99 (undercooling of 0.01) to 814 bubbles at T=0.91 (undercooling of 0.09). However

the average bubble size at the time when bubble number is maximum, which is calculated

by the total number of lattices that satisfy    c divided by the number of bubbles,

decreases with increasing undercooling. The major reason is that tmax is smaller at large

undercoolings so that the formed bubbles have less time to grow. The molecular density

distributions for systems with different temperatures at the corresponding tmax time steps

are illustrated in figure 2. It can be seen that with larger temperature undercooling the

bubbles can be denser and finer at tmax .

Converting the bubble number and waiting time to the nucleation rate is not a trivial task.

There are three reasons for this. a). The constitutional undercooling keeps changing

during system evolution at constant temperature. The system is heterogeneous throughout

the non-equilibrium processing. Secondly, the effective volume available for nucleation

shrinks due to bubble growth. Thirdly, bubbles coalesce which changes their number
density and inter-bubble spacing. These reasons prevent a comparison with classical

nucleation theory.

The effect of density fluctuation intensity on bubble formation has been considered and

the results for =0.00001 and =0.005 are illustrated in figure 3. It can be seen that

although the fluctuation affects bubble formation the effect is much weaker when

compared with that of system undercooling. The large thermodynamic driving force for

phase separation is the major contribution for bubble formation.

B. Effect of shear on bubble formation

It is interesting to address the basic question of whether shear suppresses or enhances

bubble formation. This has not been addressed clearly [6-8]. Mesoscopic interparticle

potential supplies a unique chance to tackle this question because the shear is easily

introduced in the lattice Boltzmann equation and the format of the interparticle potential

is suitable for any flow environment. For the fluid confined between two parallel plates

with the top boundary moving towards the right with velocity u x xˆ and bottom to left with

 u x xˆ , the system is initialized with a uniform molecular density. The speed at each

lattice position is 2 y / N y  1u x xˆ , where N y is the fluid depth in terms of lattice number

and y is the lattice coordinate in vertical coordinate. The results for T=0.99 and T=0.94

are presented in figures 4 and 5.

As has described earlier, the boundary shear speed of 0.05 corresponds to a shear rate in

the order of 106 s-1. The effect of shear on bubble formation, if there is any, should be

clearly evident with such a high shear rate. However, there are strong indications that the

effect of shear on the maximum number of bubbles is negligible, see figures 4(a) and

5(a). It is also noticed that Reguera and Rubi found that the effect of shear at 103 s-1 has a

negligible effect on nucleation according to their theoretical calculations based on the

Fokker-Planck equation [9]. However, figure 4(a) shows that the bubble number

decreases faster for a system under larger shear rate. The reason could be two-fold. The

shear clearly must enhance molecular transport. Therefore, the bubble grows quicker

during shear flow, as proved by figure 4(b). The faster growing of bubble reduces the

effective volume for nucleation and thus causes the reduction of new bubble formation.

Secondly, the shear enhances bubble motion and increases the opportunity for

coalescence so that the number of bubbles reduces more rapidly, as observed in the

simulation. As is well-known, shear can also cause bubbles break up in which case the

total number of bubbles might increase. However the effect is only important for large

bubbles and is beyond the scope of the present work.

Figure 5 shows that the effect of shear on bubble size and number density is much weaker

for large undercoolings. It is hard to see any difference caused by shear. The reason is

that the phase transition occurs rapidly in highly undercooled fluids. The number of

bubbles becomes maximum at 110 time steps and reduces to 20 at 2000 time steps due to

bubble coalesces. While at T=0.99, the corresponding time steps are 600 and 3500,
respectively. The effect of shear weakens as large undercoolings. Bubble coalesce plays

more important role due to the much higher bubble density.

IV. CONCLUSIONS

The main objective of this work has been the investigation of mesoscopic bubble

formation in van der Waals fluids as a function of undercooling and shear flow. The

simulations were carried out by means of the mesoscopic interparticle potentials for a

lattice Boltzmann model. It has been found that the bubble formation is accelerated and

the maximum bubble number is increased as the undercooling increased. This is expected

given the larger driving forces available at large undercoolings. The results are in

agreement with classical nucleation theory. Shear can affect bubble growth rate because

of the enhancement of mass transport but has little effect on the early stages of bubble

formation and on the maximum bubble number-density. The effect becomes negligible at

large undercoolings whence the driving force dominates.

ACKNOWLEDGMENTS

The author would like to acknowledge Professor H.K.D.H. Bhadeshia at University of

Cambridge for his careful reading and revision of the manuscript.

REFERENCES
1. N.J. Hepworth, J.W.R. Boyd, J.R.M. Hammond and J. Varley, Chem. Eng. Sci.

58, 4071 (2003).

2. I. Barbulovic-Nad, M. Lucente, Y. Sun, M.J. Zhang, A.R. Wheeler and M.

Bussmann, Crit. Rev. Biotechnol. 26, 237 (2006).

3. A. Bolognesi, C. Mercogliano, S. Yunnus, M. Civardi, D. Comoretto and A.

Turturro, Langmuir 21, 3480 (2005)

4. A. Obeidat, J. S. Li, G. Wilemski, J. Chem. Phys. 121, 9510 (2004).

5. G. Wilemski, J. S. Li, J. Chem. Phys. 121, 7821 (2004).

6. R. Blaak, S. Auer, D. Frenkel and H. Lowen, Phys. Rev. Lett. 93, 068303 (2004).

7. K. Watanabe, T. Takahashi, J. I. Takimoto and K. Koyama, J. Macromol. Sci-

Phys. 42B, 1111 (2003).

8. D. Yuya, T. Kikuchi, T. Inoue, Y. Iwai and A. Kawaguchi, J. Macromol. Sci-

Phys. 45B, 85 (2006).

9. D. Reguera and J. M. Rubi, J. Chem. Phys. 119, 9888 (2003).

10. R. S. Aga, J. R. Morris, J. J. Hoyt and M. Mendelev, Phys. Rev. Lett. 96, 245701

(2006).

11. R.D. Groot and R.B. Warren, J. Chem. Phys. 107, 4423 (1997).

12. E.G. Flekkoy and P.V. Coveney, Phys. Rev. Lett. 83, 1775 (1999).

13. X.W. Shan and H.D. Chen, Phys. Rev. E 47, 1815 (1993).

14. R.S. Qin, Phys. Rev. E 73, 066703(2006).

15. N. F. Fletcher, Eur. J. Phys. 14, 29 (1993).

16. X.Y. He and G.G. Doolen, J. Stat. Phys. 107, 309 (2002).

17. X.Y. He and L.S. Luo, Phys. Rev. E 55, R6333 (1997).
18. Q.S. Zou and X.Y. He, Phys. Fluids 9, 1591 (1997).

19. R.S. Qin and E. R. Wallach, Acta Mater., 51, 6199 (2003).
TABLE I. Bubble formation for 1024128 lattices with  = 0.005 and c=1.05. The

properties are: T, reduced temperature; beq , the equilibrium density in gas phase; t s ,

the time steps when the first bubble detected; tmax , the time steps when the number of

bubble achieves maximum; nmax , the maximum number of bubble; N , the average

number of lattices in each bubble when the number of bubbles is in maximum.

T beq ts tmax nmax N

0.99 0.844968 440 600 105 167.61

0.98 0.772406 230 335 208 122.625
0.97 0.711124 155 215 287 69.296
0.96 0.660106 120 165 379 59.15
0.95 0.617009 100 130 470 42.93
0.94 0.586988 80 110 564 40.54
0.93 0.564652 70 95 641 38.398
0.92 0.532408 60 80 723 25.9
0.91 0.497403 55 70 814 20.432
Fig. 1 Schematic diagram of shear boundaries and elementary speed vectors of d2q9
lattice Boltzmann model
Figure 2. Molecular density profiles for different temperature at tmax . (a) T=0.91; (b)

T=0.92; (c) T=0.93; (d) T=0.94; (e) T=0.95; (f) T=0.96; (g) T=0.97; (h) T=0.98 and (i)

T=0.99.
Figure 3. Effect of density fluctuation intensities on bubble formations at no-shear flow.
(a)

(b)

Figure 4. Bubble evolution at different shear rate on a system with 1024128 lattice,

initial density =2.0, defined critical density c=1.05 and T=0.99. (a). The bubble

numbers vs. time. (b). The bubble size vs. time.

(a)
(b)

Figure 5. Bubble evolution at different shear rate on a system with 1024128 lattice,

initial density =2.0, defined critical density c=1.05 and T=0.94. (a). The bubble

numbers vs. time. (b). The bubble size vs. time.